Pith. sign in

REVIEW 3 major objections 5 minor 30 references

DFR: A Decompose-Fuse-Reconstruct Framework for Multi-Modal Few-Shot Segmentation

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that a decompose-fuse-reconstruct framework that adds audio to visual and textual guidance gives the best reported few-shot segmentation results on both synthetic-audio PASCAL-5i and real-audio AVS-V3.

desk verdict First tri-modal (visual-text-audio) few-shot segmentation paper; the idea is genuinely new, results are consistent, but the synthetic audio branch is text-derived, so the audio-specific gain is not yet proven. read the letter →

arxiv 2507.16736 v1 pith:A3IMO36Q submitted 2025-07-22 cs.CV

Shuai Chen , Fanman Meng , Xiwei Zhang , Haoran Wei , Chenhao Wu , Qingbo Wu , Hongliang Li This is my paper · ORCID
classification cs.CV
keywords few-shotsegmentationmulti-modalfusionaudio-visualSegmentAnythingModelcontrastivelearningtext-guidedPASCAL-5iAVS-V3
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Few-shot segmentation asks a model to label object categories from one or a handful of labeled examples, and most systems rely on support images or on image plus a text category name. This paper claims that adding audio as a third guidance modality, through a three-stage decompose-fuse-reconstruct pipeline built on the Segment Anything Model (SAM), gives better masks than visual-only or visual-textual systems. On PASCAL-5i, using sounds synthesized from the category name by the text-to-audio model AudioLDM, the framework reports 75.4% and 76.2% mean intersection-over-union (mIoU) in 1-shot and 5-shot settings; on the real audio-visual benchmark AVS-V3 it reports 59.5% and 66.2% mIoU in 0-shot and 1-shot settings. If the claim holds, audio is a usable extra cue for few-shot segmentation, and the alignment machinery could carry over to other modalities.

What carries the argument

The load-bearing mechanism is the tri-modal decomposition that turns each modality into foreground and background features. SAM splits the support image into positive and negative region proposals by overlap with the support mask; a large language model expands the category name into descriptive attributes and background-context embeddings; and AudioLDM provides an audio embedding. An InfoNCE contrastive loss pulls anchor features (learnable foreground token, support prototype, category name, audio) toward positives (visual positive proposals, text description) and pushes them away from negatives (background token, negative proposals, co-occurring classes). The dual-path reconstruction then feeds a fused semantic token and a geometric embedding, computed from visual, textual, and audio similarity maps over query proposals, into SAM's decoder, followed by a refiner.

What would settle it

Run the same framework on PASCAL-5i with the audio embedding replaced by a fixed random vector shared by all categories, keeping the parameter count fixed: if 1-shot mIoU stays near the reported 75.4 instead of falling toward the 75.0 of Visual+Text, the audio branch is not contributing category-specific signal.

Watch

Extended reading notes

Core claim

The paper's central claim is that tri-modal guidance, in which visual, textual, and audio cues are first decomposed, then aligned by contrastive learning, and then reconstructed into SAM decoder prompts, outperforms single- and dual-modal guidance. Its reported numbers are 75.4% mIoU (1-shot) and 76.2% (5-shot) on PASCAL-5i, ahead of the visual-only SAM baselines Matcher, VRP-SAM, and FCP by 7.3, 3.5, and 2.2 points respectively in 1-shot, and 59.5% (0-shot) and 66.2% (1-shot) on AVS-V3, ahead of GAVS by 4.8 and 3.3 points. The authors interpret the consistent gains as evidence that audio carries complementary temporal-dynamic semantics rather than duplicating the text description. The discovery is method-level: a specific decomposition-fusion-reconstruction arrangement converts multi-modal prompts into measurable segmentation gains in both synthetic and real audio settings.

Load-bearing premise

On PASCAL-5i the audio is synthesized from the category name itself, so the claim that audio adds information beyond the text assumes the generated sound carries discriminative content that the text embedding does not already provide.

Editorial extensions

If this is right

  • Audio should be treated as a first-class guidance modality in few-shot segmentation, since the reported gains appear both with synthesized sounds and with real recordings.
  • SAM-based segmenters can accept multi-modal prompts without retraining the frozen image encoder; DFR only changes what is fed to the decoder and adds a contrastive alignment module.
  • The same fusion recipe transfers from a synthetic setting, where audio is generated from category names, to a real audio-visual dataset, suggesting the alignment mechanism rather than the audio source is what carries the gain.
  • Explicit negative cues, such as co-occurring classes and negative proposals, can be injected through contrastive learning to suppress background confusion in low-shot settings.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper: on PASCAL-5i the audio signal is generated as $A = \mathrm{AudioLDM}(T)$ from the same category name that produces the text branch, so the 1.7 mIoU gap between the full model and Visual+Text may partly reflect extra parameters or regularization rather than new perceptual information; the real-audio results do not settle this.
  • Beyond the paper: a controlled comparison that matches parameter counts, or that replaces the audio branch with category-independent noise, would test whether the synthetic-audio contribution is genuinely categorical; the paper's ablations do not include such a control.
  • Beyond the paper: the decompose-fuse-reconstruct recipe suggests testable extensions to other non-visual guidance such as depth or haptic cues, and to tasks like few-shot detection where proposal generation already plays an analogous role.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes DFR, a multi-modal few-shot segmentation framework that combines visual support images, textual category descriptions, and audio signals. The method has three stages: a multi-modal decomposition stage that extracts SAM-based visual proposals, LLM-expanded textual descriptors, and AudioLDM-synthesized audio embeddings; a contrastive fusion stage using InfoNCE to align anchors, positives, and negatives across modalities; and a dual-path reconstruction stage that feeds semantic tokens and geometric location priors into a SAM decoder. Experiments are reported on PASCAL-5i with synthetic audio and on AVS-V3 with real audio, claiming state-of-the-art results in both settings.

Significance. If the empirical claims are reproducible, DFR would be a useful contribution: it extends few-shot segmentation from visual-only and visual-textual paradigms to tri-modal guidance, and the decompose-fuse-reconstruct design is clearly specified with ablation tables that are directionally consistent. The use of a real audio-visual dataset as a second evaluation setting is a strength, and the paper is transparent about the synthetic-audio pipeline. However, the central claim that audio is an additive cue is not yet supported: the synthetic audio is generated from the same category name used by the text branch, no statistical significance or variance information is reported, no code is released, and the real-audio experiments lack a modality ablation and use baselines with different backbones. These issues are load-bearing for the headline claim rather than cosmetic.

major comments (3)
  1. [Sec. III-B1 / Sec. IV-A / Table III] The synthetic audio input is defined as A = AudioLDM(T), so the audio embedding f_a is a transform of the same category name T that already produces the text branch. Consequently, the 1.7 mIoU gap between the full model and Visual+Text in Table III does not establish that audio provides information beyond the text description; the gain could come from the added audio encoder, the extra InfoNCE anchor, or the modality-dropout regularization. The authors should demonstrate that synthetic audio carries independent semantics, for example by comparing against a text-irrelevant audio signal, by matching parameter counts between the compared configurations, or by adding a real-oracle audio ablation on the same benchmark.
  2. [Tables I-III] All reported results are single-run numbers without standard deviations or significance tests. The 1.7 mIoU audio gain in Table III and the 7.3-point headline gain over Matcher in Table I could fall within run-to-run variance, especially under a 10-epoch training schedule with random SAM proposal sampling and stochastic AudioLDM generation. The authors should report mean and standard deviation over at least three seeds for the main comparisons and ablations, and, where feasible, a paired significance test.
  3. [Sec. IV-C / Table II] The AVS-V3 comparison does not isolate the contribution of audio or of text from the backbone change: the baselines (AVSBench, AVSegFormer, GAVS) do not use the SAM/LanguageBind backbone of DFR, and no AVS-V3 modality ablation is reported. The claim that real audio is an additive cue therefore rests on the synthetic PASCAL-5i experiments, whose audio branch is text-derived. An AVS-V3 ablation with Visual, Visual+Text, Visual+Audio, and full tri-modal configurations, using the same backbone throughout, is needed to support the generalization claim.
minor comments (5)
  1. [Eq. (1)] There is a typo, 'wher' should be 'where', and the paper should state the overlap threshold value for τ in the text rather than only in the equation.
  2. [Eq. (3) / Sec. III-B2] The symbol f_a is used both for the anchor feature in the InfoNCE expression and for the audio embedding in the fusion module; this double use is confusing and should be disambiguated.
  3. [Eq. (5)] The similarity thresholds δ_t and δ_a are introduced but their values are never reported, even though τ in Eq. (1) and λ in Eq. (7) are specified; please provide the hyperparameter values used.
  4. [After Eq. (5)] The sentence 'δ is a similarity threshold' is repeated twice in two consecutive lines; the duplication should be removed.
  5. [General] The paper does not state whether code or trained models will be released, which would materially help reproducibility given that the central comparisons depend on a specific SAM/LanguageBind/AudioLDM pipeline.

Circularity Check

1 steps flagged · score 4.0 of 10

Synthetic audio on PASCAL-5i is generated from the same category name as the text branch (A=AudioLDM(T)), so the reported audio gain is not evidence for an independent modality.

  1. self definitional [Section III-B1 (Multi-modal Decompose, Audio Decomposition); also Section IV-A]
    "Audio Decomposition. We utilize AudioLDM [19] to synthesize characteristic sound effects A = AudioLDM(T ), which are processed to obtain audio embedding f a, providing complementary temporal-dynamic information."

    By construction the synthetic audio input A is the output of a text-to-audio model applied to the same category name T that already drives the text branch (category embedding plus LLM-generated descriptors). Thus f_a = AudioEncoder(AudioLDM(T)) is a learned transform of the text input, and it cannot supply class information independent of the text modality. The only PASCAL-5i evidence for audio's contribution is the Table III gap between Full Model (76.7) and Visual+Text (75.0); since no significance test is provided, that gap can be explained by the extra audio encoder, the additional contrastive anchor, or modality-dropout regularization rather than by independent acoustic semantics.

full rationale

DFR is an empirical systems paper and most of its comparisons are against external baselines on held-out folds, so there is no general derivation that reduces to its own assumptions. No load-bearing self-citations, no imported uniqueness theorems, and no fitted-parameter-then-prediction pattern appear. The one genuinely circular link is the synthetic audio modality on PASCAL-5i: the audio signal is defined as A = AudioLDM(T), i.e., a generated function of the same category name used by the text modality. Consequently, the audio embedding does not constitute an independent information source; the reported 1.7 mIoU gain over Visual+Text in Table III may reflect added capacity rather than complementary audio cues. The AVS-V3 experiments with real audio provide partially independent support for the central claim, but the absence of an audio ablation on that dataset leaves the synthetic-audio contribution unresolved. This is a partial circularity affecting one of the paper's two main experimental settings, so the overall score is moderate rather than severe.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rests on several unpaid inputs: frozen SAM must produce reliable region proposals, LanguageBind must align text, audio, and visual spaces, the LLM must generate usable descriptors, and on PASCAL-5i AudioLDM must create sounds that represent the category name. The paper's own learned parameters are limited to fusion layers, prompts, and a refiner, so the quality of these pretrained components is a major determinant of the reported gains.

free parameters (6)
  • overlap threshold tau (Eq. 1) = 0.5
    Manual threshold to split SAM proposals into positive and negative sets; no sensitivity analysis is reported.
  • InfoNCE temperature tau (Eq. 3) = not reported
    Temperature for contrastive loss; the same symbol is used for the overlap threshold, and the temperature value is not given.
  • similarity thresholds delta_t, delta_a (Eq. 5) = not reported
    Thresholds selecting location priors from text and audio similarities; no values or sensitivity analysis are provided.
  • loss weight lambda (Eq. 7) = 0.2
    Weight balancing segmentation losses and contrastive loss; no ablation over lambda is shown.
  • modality dropout rate = not reported
    Used during training to prevent over-reliance on specific modalities, but the dropout probability is not specified.
  • SAM proposal count N = not reported
    Number of region proposals extracted from SAM in both visual decomposition and geometric path; the value is not stated.
assumptions (5)
  • domain assumption Frozen SAM-base produces reliable region proposals and mask decoding for base and novel classes.
    The whole method relies on SAM's zero-shot region and mask quality; no fine-tuning or novel-class validation is provided beyond final mIoU.
  • domain assumption LanguageBind embeddings align visual, text, and audio spaces well enough for contrastive fusion.
    The contrastive loss and token concatenation assume a shared embedding geometry; no calibration to few-shot segmentation is shown.
  • ad hoc to paper AudioLDM generates sounds that faithfully represent the category name T.
    Used to create the entire audio modality in PASCAL-5i; no listening test or semantic verification of generated sounds is reported.
  • domain assumption DeepSeek-v3 LLM descriptions are accurate, discriminative, and free of hallucinated background classes.
    Text descriptors f_t_d and f_t_bg are generated by prompting an LLM and used without per-category verification.
  • domain assumption The Refiner module and HQ-SAM decoder behave as expected.
    Refiner(Minit, Fq) appears in Eq. 6 without architecture details; HQ-SAM decoder is referenced without citation, so its role is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DFR: A Decompose-Fuse-Reconstruct Framework for Multi-Modal Few-Shot Segmentation." pith.science (2026). https://pith.science/paper/A3IMO36Q

@misc{pith2026250716736,
  author       = {Pith},
  title        = {Pith review of: DFR: A Decompose-Fuse-Reconstruct Framework for Multi-Modal Few-Shot Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/A3IMO36Q}},
  note         = {Machine review of arXiv:2507.16736}
}
read the original abstract

This paper presents DFR (Decompose, Fuse and Reconstruct), a novel framework that addresses the fundamental challenge of effectively utilizing multi-modal guidance in few-shot segmentation (FSS). While existing approaches primarily rely on visual support samples or textual descriptions, their single or dual-modal paradigms limit exploitation of rich perceptual information available in real-world scenarios. To overcome this limitation, the proposed approach leverages the Segment Anything Model (SAM) to systematically integrate visual, textual, and audio modalities for enhanced semantic understanding. The DFR framework introduces three key innovations: 1) Multi-modal Decompose: a hierarchical decomposition scheme that extracts visual region proposals via SAM, expands textual semantics into fine-grained descriptors, and processes audio features for contextual enrichment; 2) Multi-modal Contrastive Fuse: a fusion strategy employing contrastive learning to maintain consistency across visual, textual, and audio modalities while enabling dynamic semantic interactions between foreground and background features; 3) Dual-path Reconstruct: an adaptive integration mechanism combining semantic guidance from tri-modal fused tokens with geometric cues from multi-modal location priors. Extensive experiments across visual, textual, and audio modalities under both synthetic and real settings demonstrate DFR's substantial performance improvements over state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2507.16736 by the authors.

Figure 1
Figure 1. Illustration of evolution of FSS frameworks: from visual [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the proposed Decompose-Fuse-Reconstruct (DFR) framework for multi-modal few-shot segmentation. Our approach consists of three [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Visualization of few-shot segmentation on unseen classes under [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 23 canonical work pages

  1. [1]

    Fully convolutional networks for semantic segmentation,

    J. Long, E. Shelhamer, and T. Darrell, “Fully convolutional networks for semantic segmentation,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2015, pp. 3431–3440

    2015

  2. [2]

    Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs,

    L.-C. Chen, G. Papandreou, I. Kokkinos, K. Murphy, and A. L. Yuille, “Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs,” IEEE transactions on pattern analysis and machine intelligence , vol. 40, no. 4, pp. 834–848, 2017

    2017

  3. [3]

    Ssformer: A lightweight transformer for semantic segmentation,

    W. Shi, J. Xu, and P. Gao, “Ssformer: A lightweight transformer for semantic segmentation,” in 2022 IEEE 24th international workshop on multimedia signal processing (MMSP) . IEEE, 2022, pp. 1–5

    work page 2022

  4. [4]

    Prior guided feature enrichment network for few-shot segmentation,

    Z. Tian, H. Zhao, M. Shu, Z. Yang, R. Li, and J. Jia, “Prior guided feature enrichment network for few-shot segmentation,” IEEE transactions on pattern analysis and machine intelligence, vol. 44, no. 2, pp. 1050–1065, 2020

    work page 2020

  5. [5]

    Hypercorrelation squeeze for few-shot segmentation,

    J. Min, D. Kang, and M. Cho, “Hypercorrelation squeeze for few-shot segmentation,” in Proceedings of the IEEE/CVF international conference on computer vision , 2021, pp. 6941–6952

    work page 2021

  6. [6]

    Adapt before comparison: A new perspective on cross- domain few-shot segmentation,

    J. Herzog, “Adapt before comparison: A new perspective on cross- domain few-shot segmentation,” in Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition , 2024

    work page 2024

  7. [7]

    Rethinking the correlation in few- shot segmentation: A buoys view,

    Y . Wang, R. Sun, and T. Zhang, “Rethinking the correlation in few- shot segmentation: A buoys view,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , June 2023, pp. 7183–7192

    work page 2023

  8. [8]

    Relevant intrinsic feature enhancement network for few-shot semantic segmentation,

    X. Bao, J. Qin, S. Sun, X. Wang, and Y . Zheng, “Relevant intrinsic feature enhancement network for few-shot semantic segmentation,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, 2024, pp. 765–773

    work page 2024

Show all 30 references
  1. [9]

    Label-efficient few-shot semantic segmentation with unsupervised meta-training,

    J. Li, K. Shi, G.-S. Xie, X. Liu, J. Zhang, and T. Zhou, “Label-efficient few-shot semantic segmentation with unsupervised meta-training,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, 2024, pp. 3109–3117

    2024

  2. [10]

    Image segmentation using text and image prompts,

    T. L ¨uddecke and A. Ecker, “Image segmentation using text and image prompts,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 7086–7096

    2022

  3. [11]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark et al., “Learning transferable visual models from natural language supervision,” in International conference on machine learning . PMLR, 2021, pp. 8748–8763

    2021

  4. [12]

    Extending segment anything model into auditory and temporal dimensions for audio- visual segmentation,

    J. Seon, W. Im, S. Lee, J. Lee, and S.-E. Yoon, “Extending segment anything model into auditory and temporal dimensions for audio- visual segmentation,” in 2024 IEEE International Conference on Image Processing (ICIP), 2024, pp. 2480–2486

    2024

  5. [13]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Loet al., “Segment anything,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 4015–4026

    2023

  6. [14]

    Languagebind: Extending video-language pretraining to n-modality by language-based semantic alignment,

    B. Zhu, B. Lin, M. Ning, Y . Yan, J. Cui, H. Wang, Y . Pang, W. Jiang, J. Zhang, Z. Li, C. Zhang, Z. Li, W. Liu, and L. Yuan, “Languagebind: Extending video-language pretraining to n-modality by language-based semantic alignment,” in The Twelfth International Conference on Lea...

    2024

  7. [15]

    Panet: Few-shot image semantic segmentation with prototype alignment,

    K. Wang, J. H. Liew, Y . Zou, D. Zhou, and J. Feng, “Panet: Few-shot image semantic segmentation with prototype alignment,” in proceedings of the IEEE/CVF international conference on computer vision , 2019, pp. 9197–9206

    2019

  8. [16]

    Vrp-sam: Sam with visual reference prompt,

    Y . Sun, J. Chen, S. Zhang, X. Zhang, Q. Chen, G. Zhang, E. Ding, J. Wang, and Z. Li, “Vrp-sam: Sam with visual reference prompt,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024

    2024

  9. [17]

    Matcher: Segment anything with one shot using all-purpose feature matching,

    Y . Liu, M. Zhu, H. Li, H. Chen, X. Wang, and C. Shen, “Matcher: Segment anything with one shot using all-purpose feature matching,” in The Twelfth International Conference on Learning Representations , 2024

    2024

  10. [18]

    Foreground- covering prototype generation and matching for sam-aided few-shot segmentation,

    S. Park, S. Lee, H. S. Seong, J. Yoo, and J.-P. Heo, “Foreground- covering prototype generation and matching for sam-aided few-shot segmentation,” arXiv preprint arXiv:2501.00752 , 2025

    2025 arXiv

  11. [19]

    AudioLDM: Text-to-audio generation with latent diffusion models,

    H. Liu, Z. Chen, Y . Yuan, X. Mei, X. Liu, D. Mandic, W. Wang, and M. D. Plumbley, “AudioLDM: Text-to-audio generation with latent diffusion models,” in Proceedings of the 40th International Conference on Machine Learning , vol. 202, 2023, pp. 21 450–21 474

    2023

  12. [20]

    One-shot learning for semantic segmentation,

    A. Shaban, S. Bansal, Z. Liu, I. Essa, and B. Boots, “One-shot learning for semantic segmentation,” in British Machine Vision Conference, 2017

    2017

  13. [21]

    Adaptive fss: A novel few-shot segmentation framework via prototype enhancement,

    J. Wang, J. Li, C. Chen, Y . Zhang, H. Shen, and T. Zhang, “Adaptive fss: A novel few-shot segmentation framework via prototype enhancement,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 38, 2024, pp. 5463–5471

    2024

  14. [22]

    Holistic prototype activation for few- shot segmentation,

    G. Cheng, C. Lang, and J. Han, “Holistic prototype activation for few- shot segmentation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 4, pp. 4650–4666, 2023

    2023

  15. [23]

    Few-shot segmentation via divide-and-conquer proxies,

    C. Lang, G. Cheng, B. Tu, and J. Han, “Few-shot segmentation via divide-and-conquer proxies,” International Journal of Computer Vision , vol. 132, no. 1, pp. 261–283, 2024

    2024

  16. [24]

    The pascal visual object classes (voc) challenge,

    M. Everingham, L. Van Gool, C. K. Williams, J. Winn, and A. Zisser- man, “The pascal visual object classes (voc) challenge,” International journal of computer vision , vol. 88, pp. 303–338, 2010

    2010

  17. [25]

    Semantic contours from inverse detectors,

    B. Hariharan, P. Arbel ´aez, L. Bourdev, S. Maji, and J. Malik, “Semantic contours from inverse detectors,” in 2011 international conference on computer vision. IEEE, 2011, pp. 991–998

    2011

  18. [26]

    Deepseek-v3 technical report,

    A. Liu, B. Feng, B. Xue, B. Wang, B. Wu, C. Lu, C. Zhao, C. Deng, C. Zhang, C. Ruan et al., “Deepseek-v3 technical report,” arXiv preprint arXiv:2412.19437, 2024

    2024 arXiv

  19. [27]

    Prompting segmentation with sound is generalizable audio-visual source localizer,

    Y . Wang, W. Liu, G. Li, J. Ding, D. Hu, and X. Li, “Prompting segmentation with sound is generalizable audio-visual source localizer,” in Proceedings of the AAAI Conference on Artificial Intelligence, Van- couver, Canada, 2024, pp. 5669–5677

    2024

  20. [28]

    Audio-visual segmentation,

    J. Zhou, J. Wang, J. Zhang, W. Sun, J. Zhang, S. Birchfield, D. Guo, L. Kong, M. Wang, and Y . Zhong, “Audio-visual segmentation,” in Proceedings of the European Conference on Computer Vision, Tel Aviv, Israel, 2022, pp. 386–403

    2022

  21. [29]

    Audio-visual segmentation with semantics,

    J. Zhou, X. Shen, J. Wang, J. Zhang, W. Sun, J. Zhang, S. Birchfield, D. Guo, L. Kong, M. Wang et al. , “Audio-visual segmentation with semantics,” International Journal of Computer Vision , vol. 1, 2024

    2024

  22. [30]

    Avesformer: Efficient transformer design for real-time audio-visual segmentation,

    Z. Wang, Q. Yang, L. Shi, J. Yu, Q. Liang, F. Li, and S. Xiang, “Avesformer: Efficient transformer design for real-time audio-visual segmentation,” arXiv preprint arXiv:2408.01708 , 2024

    2024 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.